基于改进Q学习的IMA系统重构蓝图生成方法

重构蓝图定义了故障状态下系统软硬件资源的重新配置方案，是实现综合模块化航空电子系统重构容错的关键。提出了一种基于改进Q学习的重构蓝图生成方法，综合考虑负载均衡、重构影响、重构时间、重构降级等多优化目标，并应用模拟退火框架改进探索策略，提高了传统Q学习算法的收敛性能。实验结果表明，与模拟退火算法、差分进化算法、传统Q学习算法相比，本文提出的改进Q学习算法效率更高，所生成重构蓝图质量更高。

Generating reconfiguration blueprints for IMA systems based on improved Q-learning

Reconfiguration blueprint defines the reconfiguration scheme of system hardware and software resources in the fault status, and is critical to reconfiguration fault tolerance of the integrated modular avionics system. In this paper, we propose an approach for generating reconfiguration blueprints based on improved Q-learning, which considers multiple optimization objectives such as load balance, reconfiguration impact, reconfiguration time, and reconfiguration degradation. The simulated annealing framework is utilized to enhance the convergence performance of the traditional Q-learning strategy. Experimental results demonstrate that compared with the simulated annealing algorithm, the differential evolution algorithm, and the traditional Q-learning algorithm, the algorithm proposed has higher efficiency, and can generate the reconfiguration blueprints of better quality.